DISCUSSION PAPER SERIES

ABCD

www.cepr.org

Available online at: www.cepr.org/pubs/dps/DP8235.asp www.ssrn.com/xxx/xxx/xxx

No. 8235

WHAT DO CEOS DO?

Oriana Bandiera, Luigi Guiso, Andrea Prat and Raffaella Sadun

FINANCIAL ECONOMICS and LABOUR ECONOMICS

ISSN 0265-8003

WHAT DO CEOS DO?

Oriana Bandiera, London School of Economics and CEPR Luigi Guiso, European University Institute and CEPR Andrea Prat, London School of Economics and CEPR

Raffaella Sadun, Harvard University

Discussion Paper No. 8235 February 2011

Centre for Economic Policy Research 77 Bastwick Street, London EC1V 3PZ, UK

Tel: (44 20) 7183 8801, Fax: (44 20) 7183 8820 Email: cepr@cepr.org, Website: www.cepr.org

This Discussion Paper is issued under the auspices of the Centre’s research programme in FINANCIAL ECONOMICS and LABOUR ECONOMICS. Any opinions expressed here are those of the author(s) and not those of the Centre for Economic Policy Research. Research disseminated by CEPR may include views on policy, but the Centre itself takes no institutional policy positions.

The Centre for Economic Policy Research was established in 1983 as an educational charity, to promote independent analysis and public discussion of open economies and the relations among them. It is pluralist and non-partisan, bringing economic research to bear on the analysis of medium- and long-run policy questions.

These Discussion Papers often represent preliminary or incomplete work, circulated to encourage discussion and comment. Citation and use of such a paper should take account of its provisional character.

Copyright: Oriana Bandiera, Luigi Guiso, Andrea Prat and Raffaella Sadun

CEPR Discussion Paper No. 8235

February 2011

ABSTRACT

What Do CEOs Do?*

We develop a methodology to collect and analyze data on CEOs' time use. The idea - sketched out in a simple theoretical set-up - is that CEO time is a scarce resource and its allocation can help us identify the firm's priorities as well as the presence of governance issues. We follow 94 CEOs of top-600 Italian firms over a pre-specified week and record the time devoted each day to different work activities. We focus on the distinction between time spent with insiders (employees of the firm) and outsiders (people not employed by the firm). Individual CEOs differ systematically in how much time they spend at work and in how much time they devote to insiders vs. outsiders. We analyze the correlation between time use, managerial effort, quality of governance and firm performance, and interpret the empirical findings within two versions of our model, one with effective and one with imperfect corporate governance. The patterns we observe are consistent with the hypothesis that time spent with outsiders is on average less beneficial to the firm and more beneficial to the CEO and that the CEO spends more time with outsiders when governance is poor.

JEL Classification: D2, G3 and G34 Keywords: CEOs, corporate governance and time use

Oriana Bandiera Department of Economics London School of Economics Houghton Street London WC2A 2AE Email: o.bandiera@lse.ac.uk For further Discussion Papers by this author see: www.cepr.org/pubs/new-dps/dplist.asp?authorid=147704

Luigi Guiso Economics Department European University Institute Villa San Paolo Via della Piazzuola 43 I-50133 Firenze ITALY Email: luigi.guiso@eui.eu For further Discussion Papers by this author see: www.cepr.org/pubs/new-dps/dplist.asp?authorid=104028

Andrea Prat STICERD London School of Economics Houghton Street London WC2 2AE Email: a.prat@lse.ac.uk For further Discussion Papers by this author see: www.cepr.org/pubs/new-dps/dplist.asp?authorid=138721

Raffaella Sadun Harvard University 100 Edwin H Land Boulevard Cambridge, MA 02138 USA Email: For further Discussion Papers by this author see: www.cepr.org/pubs/new-dps/dplist.asp?authorid=164835

* We thank seminar participants at Catholic University of Milan, Columbia University, EEA Meetings, Harvard Business School, Imperial College, IMT Lucca, LSE, MIT, New York University, NIESR, Universitat Pompeu Fabra, Stockholm (IIES), and the University of Sussex for useful comments. We are grateful to Tito Boeri, Daniel Ferreira, Luis Garicano, Steve Pischke, Imran Rasul, and Fabiano Schivardi for useful suggestions. This research was made possible by generous funding from Fondazione Rodolfo Debenedetti.

Submitted 29 January 2011

1 Introduction

Time management has been at the core of the research on management for almost �fty years.

Peter Drucker (1966) makes it the centerpiece of his celebrated book on e¤ective management:

�E¤ective executives know that time is the limiting factor. The output of any process is set

by the scarcest resource. In the process we call �accomplishment�, this is time.�

More recently, the importance of time constraints for people at the top of organiza-

tional hierarchies has been recognized by a number of economic models, like Radner and Van

Zandt�s work on hierarchies (surveyed in Van Zandt, 1998), Bolton and Dewatripont (1994),

and Garicano (2000). Intellectual tasks �information processing in Radner-Van Zandt, com-

munication in Bolton-Dewatripont, problem solving in Garicano �require managerial time,

which is particularly scarce at the top of the organization. As a consequence, a key question

in organization economics is whether and how managers allocate their time to maximize �rm

performance. The theoretical and practical interest, however, are not matched by adequate

empirical evidence, as �rm datasets typically contain no information on managerial time use,

while existing studies of managerial time use are mostly based on very few observations and

are not matched to �rm level outcomes.1

This paper makes a step towards �lling this gap by developing a methodology to system-

atically collect information on how CEOs allocate their time between di¤erent work activities.

We develop a time-use diary for CEOs and we use it to record the time allocation of a sample

of 94 CEOs belonging to leading Italian companies in various industries over a pre-selected

work week. A minimalistic model of managerial use of time guides our interpretation of ob-

served di¤erences in time allocation patterns, and how these are related to the �rms�priorities

and to di¤erences between the �rms�and the CEOs�interests.

To collect the time use data, we ask the CEO�s personal assistant (PA) to keep a diary of

the activities performed by the CEO during a pre-speci�ed week (from Monday to Friday).

The PA has full information on the CEO�s schedule and he/she has a good understanding of

the functions of the people that his/her boss interacts with.2 The PA reports overall working

time and records in a diary all the activities of the CEO that last longer that 15 minutes

detailing their characteristics, in particular information on other participants.

To operationalize the concept of time allocation, we group activities according to whether

1For example, Mintzberg (1973) seminal work of managerial time use is based on a sample of 5 CEOs. To

the best of our knowledge, the largest observational dataset on top executives is Kotter (1999), which includes

15 general managers, who were selected for being �successful�.2Roxanne Sadowski, Jack Welch�s long-term PA, makes this abundantly clear: �For more than fourteen

years, I�ve been human answering machine, auto dialer, word processor, �ltering system, and fact checker;

been a sounding board, schlepper, buddy, and bearer of good and bad tidings; served as a scold, diplomat,

repairperson, cheerleader and naysayer; and performed dozens of other roles under the title of �assistant�.�

2

they involve employees of the �rm (insiders) or only people external to the �rm (outsiders).

Insiders are listed by functional area, for instance �nance, marketing, or human resources.

Outsiders are grouped in categories CEOs normally interact with, for instance suppliers,

investors, or consultants. The insider/outsider classi�cation is ideal for our purposes for two

reasons. First, it captures the essence of the CEO�s job: �The CEO is the link between the

Inside that is the organization, and the Outside of society, economy, technology, markets, and

customers �(Drucker, cited in La�ey 2009). Second, it is a dimension over which the CEO�s

and the �rm�s interests might be misaligned. As a matter of fact, the optimal inside/outside

balance is the subject of controversy in the management, �nance and economics literature. At

one end of the spectrum, a widely held view is that since the CEO is the "public face" of the

company, spending time outside the �rm is an important (if perhaps not the most important)

role of the CEO. 3 At the other end, however, an increasingly popular view points out that

time spent with outsiders might mostly bene�t the CEO without contributing to creating

value for the �rm. Khurana (2002), for instance, argues that CEOs have strong incentives

to seek visibility, by cultivating personal connections with in�uential business leaders. In

line with this, Malmendier and Tate (2009) show that when their power vis-a-vis the �rm

increases, CEOs spend more time outside the �rm in activities, such as writing books and

playing golf, and that this shift in activities does not contribute to �rm�s performance.4

Our data reveal that, as expected, CEOs spend the majority of their time with other

people (85%). Of these most are employees of the same �rm, but many are not. On average,

CEOs spend 42% of their time with insiders only, 25% with both insiders and outsiders and

16% with outsiders alone. More interestingly, these averages hide a great deal of heterogeneity,

both in terms of number of hours worked and time allocation. A majority of CEOs spend

very little time (less than 5 hours per week) alone with outsiders, but over 10% of them

spend over 10 hours a week. Our empirical analysis is aimed at making sense of the wide

heterogeneity on this dimension of time allocation choices.

Since the implication of di¤erent time allocation choices - and in particular the di¤er-

ential impact of time spent with insiders vs. outsiders - is a priori ambiguous, we develop

a minimalistic model of managerial time allocation that allows us to distinguish productive

activities from activities that mostly confer private bene�ts to the CEO. We use this frame-

3Procter & Gamble�s CEO A. G. La�ey (2009) views the link with the outside as the raison d�être of the

top executive: �The CEO alone experiences the meaningful outside at an enterprise level and is responsible

for understanding it, interpreting it, advocating for it, and presenting it so that the company can respond in

a way that enables sustainable sales, pro�t, and total shareholder return (TSR) growth.�4Khurana (2002)�s view is that the market for CEOs is deeply �awed. The search for external, charismatic

�corporate saviors�leads boards �and the executive search companies they employ �to give their preference

to highly visible personalities that can more easily be "sold" to the �rm shareholders. In turn, this creates an

incentive for CEOs to seek visibility, by cultivating personal connections with in�uential business leaders.

3

work to shed light on whether time with insiders and outsiders can be interpreted along these

lines. In our model, the CEO chooses how much time to devote to a number of possible

work-related activities, or to leisure. Each of the work activities yield non-negative bene�ts

to the �rm and to the CEO. For instance, networking with clients might increase the �rm�s

sale but also increase the chance that the CEO is o¤ered a better job in the future. Suppose

activities can be roughly divided into the ones that bene�t mostly the �rm and those that

bene�t mostly the CEO. In this set-up, the interpretation of di¤erences in time allocation

between these two sets of activities depends on the extent to which the CEO�s and the �rm�s

interests are aligned. When these are perfectly aligned, time allocation is entirely determined

by the �rm�s objectives. In this case, the observed variation in time allocation is entirely de-

termined by optimal responses to di¤erent conditions faced by di¤erent �rms. However, when

the CEO�s and the �rm�s interests are not perfectly aligned, time allocation is determined by

both the �rm�s objectives and the CEO�s objectives. In this case, the observed variation in

time allocation partially re�ects di¤erences in the quality of governance that determines the

alignment of the CEO�s interests with the �rm�s.

If the observed variation re�ects di¤erences in governance, the model makes precise that

the following three sets of correlations must hold:

1. CEOs who work longer hours, devote more time to activities that mostly bene�t the

�rm and less time to activities that mostly yield private bene�ts.

2. CEOs who work for �rms with stronger governance devote more time to activities that

mostly bene�t the �rm and less time to activities that mostly yield private bene�ts.

3. Time devoted to activities that mostly bene�t the �rm is more strongly correlated with

productivity than time devoted to activities that mostly yield private bene�ts.

Guided by the theoretical framework, we estimate the correlation between the CEOs�time

use, their total time at work, �rm level measures of governance and productivity, conditional

on the size of the �rm, whether it is listed and the industry it is in. The analysis yields three

key �ndings, corresponding to results 1-3 above.

First, the insider/outsider allocation is associated with systematic di¤erences in CEO

worktime. In particular, CEOs who work longer hours spend more time with insiders and

less time, in absolute terms, with outsiders, especially in on one-on-one meetings.

Second, the insider/outsider allocation is systematically correlated with di¤erences in

�rm governance. More precisely, we correlate CEOs time allocation with a range of external

proxies for �rm level governance. We begin by comparing the time allocation of CEOs working

for domestic and multinational �rms. We expect the latter to have stronger governance as

4

multinational �rms face global competition and therefore need to provide steeper incentives

to maintain their productivity advantage (Bloom and Van Reenen 2010 and Bandiera et. al

2010). We then look at variables that proxy for the CEO�s power vis-a-vis the �rm owners.

First, we compare external CEOs who work for family �rms to those hired by �rms owned

by disperse shareholders. Intuitively, when ownership is concentrated, as it is in family �rms,

owners might face less di¢ culties in monitoring the CEOs�time use. Second, we test whether

time allocation is correlated with board size. Large boards have long been suspected of being

dysfunctional and to lead to too little criticism of management performance, thus granting

CEOs more freedom (e.g. Lipton and Lorsch, 1992). Consistent with this view, there is

some evidence that larger boards reduce the value of the �rm (e.g. Yermack, 1996). Third,

following a recent literature pointing at the fact that the presence of women on the board

might improve the quality of governance (Adams and Ferreira 2009), we use an indicator

for whether there is at least a woman in the board with an executive role. Finally, as a

more direct (and arguably exogenous) measure of governance quality and restraints on the

managers we use the country-level indicator of private bene�ts of control computed by Dyck

and Zingales (2004) on the basis of block control transactions in 39 countries. We attach

this measure of quality of governance to multinational �rms in our sample on the basis of

their country of origin and rely on within variation among multinational �rms operating in

Italy. The correlations between time use and the �ve independent governance proxies paint

a consistent picture: CEOs who work for �rms with better governance devote more time to

insiders and less time to meeting outsiders alone.

In the �nal part of our empirical analysis, we show that the insider/outsider allocation

is correlated with �rm performance. Time spent with insiders is positively correlated with

several measures of �rm performance, while time spent with outsiders only is not. For

example, a one percent increase in hours dedicated to insiders is correlated with a productivity

increase of 1.22%, whereas a one percent increase in hours dedicated to outsiders is correlated

with a productivity increase of 0.22%, and both e¤ects are signi�cantly di¤erent from zero

at conventional levels.

Taken together, our �ndings are consistent with the idea that di¤erences in time allocation

re�ect di¤erences in governance and that, compared to time spent with outsiders alone, time

spent with insiders is more bene�cial to the �rm. Since the data at hand, however, is not

suited to identify the causal e¤ect of governance on time allocation, we can make precise the

set of assumptions needed to reconcile the �ndings with the view that di¤erences in time

allocation re�ect optimal responses to di¤erent shocks faced by di¤erent �rms. The �nal

section of the paper discusses these in detail.

The paper is structured as follows. After a brief literature review in Section 2, Section

5

3 explains our empirical methodology and describes the data. Section 4 presents a simple

time allocation model. Section 5 reports empirical �ndings based on our theoretical set-up.

Section 6 concludes.

2 Related Literature

The collection and analysis of time use data has experienced a recent surge in the economics

literature, mostly to shed light on the work/leisure choice and to measure welfare (see, e.g.

Aguiar and Hurst 2007, Krueger et al 2009). Closer to our exercise are Hamermesh (1990)

and Luthans (1988), who analyse how workers and middle managers, respectively, spend time

at work. Hamermesh (1990) analyzes time diaries of a sample of 343 US workers to shed light

whether time devoted to leisure on-the-job (e.g. on breaks) is productive for the �rm, or an

indicator of shirking. Luthans (1988) records in detail the activities of 44 managers with

the aim to identify the determinants of a manager�s success within his or her organization.

Other celebrated management authors, such as Kotter or Mintzberg, also base their works on

in-depth observation of managerial behavior, however they tend to base their research on in

depth observations of a limited number of managers. Although this literature has provided a

very rich set of information on the nature of managerial activities, it has not directly tested

the implications of di¤erent time use allocations for �rm performance. To the best of our

knowledge, this is the �rst paper that measures how CEOs spend their time using a large

sample of individuals.

Our analysis is also related to the vast literature that focuses on agency problems for CEOs

(see the survey by Stein (2003)). Our paper studies one particular form of moral hazard,

that relates to the diversion of CEOs time allocation, and hence it is quite far from other

types of misbehavior studied in the literature. The most closely related study is Malmendier

and Tate (2009), who look at what happens when CEOs receive prestigious business awards,

both in terms of �rm performance and CEO behavior. One of their �ndings is that award

recipients appear to devote more time to writing books, playing golf, and sitting on boards

of other companies, a �nding that is not inconsistent with ours in so far time to write books

crowds CEO�s out working time rather than his leisure.

Our paper is complementary to Bertrand and Schoar�s (2003) analysis of management

style in a panel of matched CEO-�rm observations. While their objective is to identify

the �xed e¤ect of individual managers, our goal is to use a direct behavior variable - time

allocation - to make inferences on the implicit incentive structure that managers face.

6

3 The CEOs�Diaries: Descriptive Evidence on Time Use

3.1 Sources and Sample Description

Our main data source is a time use survey, which we designed to identify the activities CEOs

engage in on a day-to-day basis. The survey e¤ectively shadows the CEO through his PA

for every day over a one week period. The PA is asked to record real-time information on

all the CEO�s activities that last 15 minutes or longer in a time use diary. Table A1 shows

how the diary is structured. For each activity - de�ned as a task to which the CEOs devotes

time in excess of 15 minutes - the diary records information on the type of activity (e.g.

meetings, phone calls etc), its duration, its location, whether it was scheduled in advance

and when, whether it is held regularly and how often. The diary also collects information

on the number of participants, whether these belong to the �rm or not, and if insiders,

their occupational areas (e.g. �nance, marketing); if outsiders, their relation to the �rm

(e.g. investors, suppliers). The PA is also asked to record the total time the CEO spends

in activities that last 15 minutes or less and in transit. Hence, by summing the time spent

over activities in excess of 15 minutes and the time spent in activities that last less than 15

minutes we obtain a measure of the CEO total working time. Each PA is randomly assigned

one of �ve weeks between February 10 and March 14, 2007.

The master sample contains the top 800 Italian �rms from the Dun&Bradstreet data

base (2007) and the top 50 Italian banks from the list of all major Italian �nancial groups

compiled yearly by the Research Division of Mediobanca, a leading Italian investment bank.

Size is measured as yearly revenue for �rms and as the average of (i) employment, (ii) Stock

market capitalization; (iii) Total value of loan portfolio for banks. All 850 institutions were

contacted to ascertain the identity and contact details of the CEO; this procedure yielded 720

complete records. Of these, 50 were randomly selected for a pilot survey and the remaining

670 formed the �nal sample. The response rate was 18%, yielding complete information on

the time use of 119 CEOs.

The implementation of the survey was outsourced to a professional survey �rm, Carlo

Erminero and co., headquartered in Milan. Sample CEOs received an o¢ cial invitation letter

from the Fondazione Rodolfo Debenedetti that sponsored this project, followed by a personal

phone call explaining the purpose of the survey and the relevant con�dentiality clauses. Upon

acceptance, the survey was mailed to the PA identi�ed by the CEO, who was asked to record

the information and send back the completed forms via either fax or mail.

To maintain comparability across individuals, we drop ten CEOs who also cover the role

of "board president". The �nal sample contains 94 CEOs. 5

5We analyzed whether the survey sample was systematically di¤erent from non respondents in terms of

7

We match the time use data with two further sources. The �rst is the Amadeus data

base, which contains balance sheet data, �rm size, ownership and multinational status for

94 �rms in our sample. The second is the Infocamere database, which provides a rich set of

demographic information about all board members with executive and non executive respon-

sibilities.

Our dataset has obvious strengths and appeals; but it is subject also to a number of

limitations which deserve to be mentioned up-front. Apart from its small size, which re�ects

our focus on the largest corporations together with a low survey response rate (18%), the

data are cross-sectional in nature. We do not observe work-related activities that the CEO

may perform in the evenings or on weekends, and we miss the characteristics of those that

take less than 15 minutes.

Table 1 describes our sample �rms and CEOs. Panel 1A shows that manufacturing is the

most common sector (39%) followed by Finance, Insurance and Real Estate (15%), Services

(14%), Transportation, Communication and Utilities (13%), Wholesale and Retail (8%),

Construction (4%) and Mining (3%). The median �rm in our sample has 1,244 employees

and its productivity (measured as value of sales per employee) is USD 600,000 per year. Just

under 40% of the sample �rms are widely held, 22% belong to families, and the remainder is

split between private individuals, private equity, government and cooperatives. Finally, the

sample 30% of the �rms are domestic, 36% Italian multinationals and 34% Italian subsidiaries

of foreign multinationals. For the latter, respondents in our sample are CEOs of the Italian

subsidiary.

3.2 CEOs�Time Use: Insiders vs Outsiders

Table 2 and Figure 1 illustrate how CEOs spend their time. The average CEO in our sample

works 47.7 hours over a 5 day-week. We note that this only comprises the work hours known

to the PA, and as such it does not include working time at home or at the weekend. The

�gure thus provides a lower bound on the working hours of the CEOs. Of these, 36.9 -just

under 80%- are spent in activities that last 15 minutes or longer, for which the diary records

detailed information. The rest of the analysis will focus on these activities that, reassuringly,

comprise the bulk of the CEO�s time.

observable �rm level characteristics, such as size, productivity, pro�ts, location and industry. The analysis

(presented in Table A2 in appendix) shows that the CEOs participating in our study tended to work for larger

�rms, and had a higher probability of being a¢ liated to the Fondazione Debenedetti. In terms of industry

representativeness, we also had a higher response rate for �rms classi�ed in Public Administration, Business

Services and Finance. However, we could not �nd any systematic di¤erence in terms of regional location

and �rm performance, as measured by productivity, pro�ts and three years sales growth rates. The selection

analysis is limited to the sample of 535 �rms that could be matched to the Amadeus dataset

8

Table 2 (top panel) and Figure 1 show that, as expected, CEOs spend most of their time

(85%) with other people. Meetings take up 60% of the working hours, and the remaining

25% is comprised of phone-calls, conference calls and public events. Our sample CEOs

spend on average 15% of their time working alone. We note that this might be subject to

measurement error, as the CEO might be in his o¢ ce but not necessarily working. This is

of little consequence for our analysis, which focusses on the time the CEO spends with other

people.

We illustrate how the CEOs allocate their time across di¤erent categories of people, and

we focus especially on the distinction between people who belong to the �rm (insiders) and

those who do not (outsiders). The diaries reveal three interesting patterns. First, while

insiders are present most of the time, CEOs also spend a considerable amount of time with

outsiders alone (Figure 1B). On average, CEOs spend 42% of their time with insiders only,

25% with both insiders and outsiders and 16% with outsiders alone. Second, time spent with

insiders is evenly spread across di¤erent operational areas (Table 2B). Among the top �ve

categories of insiders -ranked by total time spent, the ratio between the highest (�nance: 8.6

hours) and the lowest (human resources: 5.5 hours) is 1.4. In contrast, consultants dominate

among outsiders. Looking again at the top �ve categories by time spent, the ratio between

the highest (consultants: 4.7 hours) and the lowest (suppliers: 1.3 hours) is 3.9. Third,

most of the variation between insiders is on the intensive margin: at least 70% of our sample

CEOs spends at least 15 minutes per week with each of the top �ve categories of insiders. In

contrast, time spent with di¤erent categories of outsiders exhibit considerable variation on

the extensive margin.

More interestingly, we note that the average values hide a considerable amount of vari-

ation. Figure 2 shows the histograms of total diary-recorded hours, and hours spent with

insiders and outsiders. There is considerable variation in hours worked: the CEO at the

90th percentile works 20 hours longer than the CEO at the 10th percentile (47 vs 27). The

di¤erences in hours spent with various participants are even starker. The CEO at the 90th

percentile devotes 35 weekly hours to activities where there is at least one insider, whereas

the corresponding �gure is 14 hours for the CEO at the 10th percentile. The CEO at the

90th percentile devotes 11 weekly hours meeting alone with outsiders, whereas the CEO at

the 10th percentile spends no time at all with outsiders alone.

To test whether these di¤erences are systematic rather than the outcome of random

variation across CEOs in time input requirement or purely driven by �rm characteristics, we

estimate the time spent with at least one insider at the most disaggregated level of observation

in our data - the activity level - conditional on �rm size, industry dummies and CEO �xed

e¤ects. We observed a total of 2,612 activities in the sample and on average, CEOs engage

9

in 27 activities over the observation week. We reject the null hypothesis of no systematic

di¤erence, namely that the CEO �xed e¤ects are jointly equal to zero, at the 1% level. The

estimated �xed e¤ects, reported in Figure A1 with their 95% con�dence interval, illustrate

that conditional on �rm size and industry, CEOs exhibit di¤erent patterns of time use.

4 A Model of CEO Time Allocation

The goal of this section is to develop a minimalistic model of managerial time allocation

to illustrate how the interpretation of di¤erences in time allocation among di¤erent CEOs

depends on the extent to which the CEO�s and the �rm�s interests are aligned. We will show

that when these are perfectly aligned, time allocation is entirely determined by the �rm�s

objectives, whereas when they are not, time allocation re�ects di¤erence in governance. In

this case the model will produce a set of testable correlations that can be tested with our

data. The points we make would still hold in a more general model, but �in the interest of

space and readability �we prefer to illustrate them in the simplest possible linear quadratic

formulation.

The section ends by dealing with the connection between desired time allocation, which

is the activity time distribution predicted by the model, and observed time allocation, which

is the particular realization of the theoretical time distribution that we observe in our CEO

diaries.

4.1 De�nitions

The CEO faces n activities and allocates non-negative time vector (x1; :::; xI) to the activities.

The �rm�s production function is

Y =

nXi=1

�ixi

The vector � describes the value of the top manager�s time in all possible activities and it is

determined by the �rm technology and environment.

The CEO can also produce some personal rent (e.g. networking), with production function

R =

nXi=1

�ixi

The vector � depends on characteristics of the CEO and the institutional and economic

environment he operates in.

10

The total cost of time for the CEO is

C =1

2

nXi=1

x2i :

There is an increasing marginal cost in devoting time to one particular activity, due either

to the onset of boredom or to a lower time-e¢ ciency (for instance, take the time devoted to

having lunch with clients; the time spent to meet one client equals one hour for the meal plus

transportation time; the CEO �rst meets the clients who are willing to come to the �rm or

those who work nearby; then he has to spend more time traveling if he wants to meet the

ones in more distant locations).6

The CEO�s payo¤ is

u = bY + (1� b)R� C:

The parameter b captures the alignment between the �rm�s interests and the CEO�s �implicit

or explicit � incentive structure. If b = 1, the �rm and the CEO have perfectly aligned

interests. If b = 0, the CEO only pursues personal interest.

4.2 Perfect Governance

We begin by analyzing the benchmark case where the CEO�s and the �rm�s interests are

perfectly aligned The CEO solves

maxx

nXi=1

�ixi �1

2

nXi=1

x2i

and the solution is:

Proposition 1 In the perfect-alignment case (b = 1), the CEO devotes time to activities in

proportion to the relative value of the activities to the �rm. Namely, given activities i and j,

the equilibrium allocation of time on the two activities (xi and xj) satis�es:

xixj=�i�j:

6This formulation is equivalent to one where the production function is logarithmic

Y =

nXi=1

�i log xi

R =

nXi=1

�i log xi

and the cost function is linear

C =

nXi=1

xi:

11

The proposition makes clear that when the CEO�s interest is perfectly aligned with the

�rm�s, observing data on time allocation allows us to back out the relative importance of the

di¤erent activities in the eyes of the �rm.

4.3 Imperfect Governance

When b 2 (0; 1), the CEO pursues both the �rm�s interest and her own. The solution to theCEO�s maximization problem no longer implies that time is allocated according to the �rm�s

needs. Given two activities i and j, we have:

xixj=b�i + (1� b) �ib�j + (1� b) �j

:

Now, with no additional information, we are unable to back out the coe¢ cients � from

the allocations x for a particular CEO. However, we can derive a number of cross-sectional

implications that will guide our empirical work.

To put some structure on the problem, We make two assumptions. First, we assume that

activities can be grouped into two sets: IY and IR. The �rst set contains activities that

bene�t the �rm more than the CEOs (�i > �i when i 2 IY ), while the second set containsactivities that yield more personal bene�t (�i < �i when i 2 IR). Furthermore, activities inIY are better for the �rm: if i 2 IY and j 2 IR, then �i > �j (otherwise there are activitiesthat are dominated). In particular, our assumption is satis�ed when activities can be split

into two groups: tasks that only bene�t the �rm (�i > 0 and �i = 0) and tasks that only

bene�t the CEO (�i = 0 and �i > 0). As shorthand, we call activities in IY �bene�cial to

the �rm�and activities in IR �bene�cial to the CEO,�but recall that this de�nition should

be preceded by �comparatively more�.

Second, we assume that the bene�t to the �rm from Y activities is greater in monetary

terms than the bene�t to the CEO of R activities, namelyXk2IY

�k >Xk2IR

�k

This is a weak condition; its violation would imply that the CEO job exists mainly for the

bene�t of the job holder rather than for the �rm.

There is a distribution of �rms that are identical except for the quality of governance, so

that b is a random variable with support (0; 1). If we observe a distribution of realized time

allocations (a vector x for each �rm), we can state

Proposition 2 In equilibrium, the cross-sectional correlation between the time xi that theCEO devotes to a particular activity and the total time the CEO spends at work is positive

12

for activities that are bene�cial to the �rm and negative for activities that are bene�cial to

the CEO.

Corr

nXk=1

xk; xi

!> 0 () i 2 IY

Proof. Recall that the equilibrium allocation is given by xi = b�i+(1� b) �i. The covariance�whose sign is the same as the correlation �is

Cov

nXk=1

xk; xi

!= Cov

nXk=1

(b�k + (1� b) �k) ; b�i + (1� b) �i

!

= V ar (b)

�i

nXk=1

�k � �inXk=1

�k � �inXk=1

�k + �i

nXk=1

�k

!

= V ar (b) (�i � �i)

nXk=1

�k �nXk=1

�k

!

which is positive if and only if �i > �i.

The intuition behind proposition 2 has to with better discipline. Recall that �rms di¤er

in terms of governance quality b. Firms with a higher b induce the CEO to spend more

time on Y -tasks that bene�t the �rm (relatively more) and less on R-tasks that bene�t the

CEO. This determines an increase and a decrease in total work time. However, the second

assumption guarantees that the former e¤ect is strictly larger. That�s because there is more

potential bene�t to be gained through activities that are bene�cial to the �rm. If the �rm

owners are able to translate this into monetary incentives for the CEO, they can obtain a

higher level of e¤ort than the CEO is willing to put on activities that are directly bene�cial

to herself.

Second, suppose we have a, possibly noisy, measure of governance. Let

b = b+ ";

where b is the governance proxy and " is an i.i.d. noise term. Note that b could be a poor

proxy, in the sense that the variance of " can be very high.

Unsurprisingly, better governance is positively related to activities that are more bene�cial

to the �rm and negatively related to R activities:

Proposition 3 The governance measure b is positively correlated with time spent on anactivity if and only that activity is bene�cial to the �rm.

Cov�b; xi

�> 0 () i 2 IY ;

13

Proof. We have

Cov�b; xi

�= Cov (b+ "; b�i + (1� b) �i) = V ar (b) (�i � �i) ;

whose sign depends �once more �on whether i is a productive activity.

Finally, we can relate time allocation to �rm productivity Y .

Proposition 4 In equilibrium, the cross-sectional correlation between the time xi that theCEO devotes to an activity i and �rm�s productivity Y is positive if and only if activity i is

bene�cial to the �rm. Namely

Corr�Y ; xi

�> 0 () i 2 IY ;

Proof. Recall that

Y =nXk=1

�kxk =Xk2IY

�kxk

We have

Cov�Y ; xi

�= Cov

Xk

�k (b�k + (1� b) �k) ; b�i + (1� b) �i

!

= Cov

Xk

�b�2k + (1� b)�k�k

�; b�i + (1� b) �i

!

= V ar (b)

�i

nXk=1

�2k � �inXk=1

�2k � �iXk

�k�k + �iXk

�k�k

!

= V ar (b) (�i � �i)

nXk=1

�k (�k � �k)!

which � again � is positive if and only �i > �i, because there exists �� such that �i > ��

whenever i 2 IY and j 2 IR and hencePnk=1 �k (�k � �k) �

Pnk=1 �� (�k � �k) � 0.

The usefulness of the three previous results is that they allow us to identify the ralative

value of an activity to the �rm and the CEO. If we do not know a priori which activities are

more bene�cial to the �rm and which activities are more bene�cial to the CEOs, the three

propositions o¤er three distinct ways of using time allocation information to distinguish

between these. In the empirical application we use this framework to shed light on whether

time with insiders and outsiders can be divided along these lines.

5 Evidence on CEOs�Time Use

Guided by the theoretical framework we now explore the correlation between the time use

of the CEO and three sets of variables: (i) CEOs�e¤ort, (ii) �rm governance and (iii) �rm

14

performance. It is key to stress that our analysis only uncovers equilibrium correlations, that

is the nature of the data does not allow us to establish causal links. The analysis will however

shed light on the set of assumptions under which the evidence can be made consistent with

the two views of the world -perfect vs imperfect governance- formalized by the theoretical

framework. We will discuss these assumptions in detail in Section 6.

5.1 CEOs�Time Use and Hours Worked

The descriptive evidence in Figure 2 indicates that CEOs di¤er considerably both in the

amount of time they work and in how they allocate this across di¤erent categories of people.

Our �rst test aims at assessing whether e¤ort -measured by hours worked- and time use are

correlated. Namely do CEOs who work longer hours have a systematically di¤erent pattern

of time allocation between insiders and outsiders?

Throughout we focus on the distinction between time spent with at least one insider- with

or without outsiders- and time spent with outsiders alone. This follows from the reasonable

assumption that if the CEO devotes some of his time to activities that only carry private

bene�ts, he is more likely to do so when no other �rm employee is present.

Table 3 estimates the correlations between hours worked and time allocated to insiders

and outsiders alone. Throughout we control for �rm size, whether the �rm is listed and

industry dummies (coe¢ cients not reported), as the time CEOs must devote to outsiders

is likely to be determined by these. Panel A shows that there is a signi�cant correlation

between the total time the CEO spends at work and how he allocates this between di¤erent

categories of people. In particular, hours devoted to insiders increase one to one with hours

worked, whereas hours spent with outsiders alone are negatively correlated with total hours

worked. The table reports the p-value of the test of the null hypothesis that the coe¢ cient is

equal to one- namely that hours with insiders (outsiders) increase in direct proportion to the

hours worked. We �nd that the coe¢ cient in the regression of the logarithm of hours spent

with insiders on the logarithm of total hours is not statistically di¤erent from one, whereas

the coe¢ cient in the corresponding regression for hours spent with outsiders only is negative,

and signi�cantly di¤erent from one.

Panel B shows that the negative correlation between hours worked and hours spent with

outsiders alone is driven exclusively by the fact that CEOs who work longer hours devote

fewer hours to one-to-one meetings with outsiders alone. Columns 3 to 6 divide the time spent

with insiders and outsiders into time spent with one or more categories of each. For instance,

a meeting with investors would fall under the "one category of outsiders" heading, whereas

a meeting with investors and consultants would fall under the "more than one category of

outsiders" heading. Columns 3 and 4 show that the correlation between hours worked and

15

time spent with insiders does not depend on whether these belong to one or more categories.

In contrast, Columns 5 and 6 show that CEOs who work longer hours spend fewer hours in

meetings alone with one type of outsiders only. The correlation with hours spent with more

than one category of outsiders is positive but signi�cantly less than one.

Panel C divides the time spent with insiders and outsiders according to the number of

people participating in the activity. In line with the �ndings above, Columns 7 and 8 show

that the correlation between hours worked and time spent with insiders does not depend on

the number of insiders present. In contrast, Column 9 shows that CEOs who work longer

hours spend fewer hours in meetings with one or two outsiders and nobody else.

5.2 CEO�s Time Use and Governance

Findings so far indicate that CEOs who work longer hours devote more time to insiders and

less time- in absolute terms- to meeting outsiders alone, especially in one-to-one meetings. To

the extent that CEOs whose �rms have better governance work harder, this is consistent with

the interpretation that observed variation in time allocation re�ect di¤erences in governance.

We can test this relationship directly, by analyzing the correlation between time use and �ve

independent proxies of "governance quality". These proxies rely on the intuition that the

CEO�s ability to pursue activities that only produce private bene�ts, depends on the strength

and e¤ective functioning of the �rm board and its ability and willingness to monitor CEO�s

actions, as well as on the economic cost the CEO pays if the �rm underperforms.

In this framework, governance quality is both the re�ection of economic incentives and

the board�s ability to monitor the CEO. Since misalignment between the �rm and the CEO

objectives is not directly observable, we proxy quality of governance using �ve external �rm

level indicators. First we use an indicator of whether the �rm is a multinational. Intuitively,

because of the stronger competitive pressure that multinationals face, they need keep CEOs

focused on �rm productivity to preserve their competitive advantage. In line with this Bloom

and Van Reenen (2010) and Bandiera et. al (2010) show that multinational �rms are more

likely to be well managed and to o¤er steep performance incentives to their executives. Hence

we would expect that CEOs of multinational �rms incur a higher cost if they divert time

from productive to unproductive uses. Supporting this idea, we �nd that CEOs who work

for multinationals spend more time with insiders and less time with outsiders alone. In

particular, Columns 1 and 2 in Table 4 show that CEOs who work for foreign multinationals

spend 54% more time with insiders and 55% less with outsiders alone. Both coe¢ cients are

precisely estimated at conventional levels. Time allocation in Italian multinationals is similar

and we cannot reject the null that the coe¢ cients are equal.

The next set of indicators proxies for the power of �rm owners vis-à-vis the CEO. The

16

�rst indicator in this set measures whether the �rm is owned by a family, and if so, whether

the CEO is external to the family or one of its members. This distinction is key because

it raises di¤erent governance issues. Compared to disperse shareholders, family owners face

fewer di¢ culties in monitoring the CEOs they hire to run their �rms, for instance because

coordination among owners is much easier when owners are few. When the CEO belongs to

the family, there is no separation of ownership and control and hence no governance issues

as such; however, family CEOs in family �rms might pursue objectives other than pro�t

maximization (Bandiera et al 2010) and the implication of these for time allocation is not

captured by our simple framework. Columns 3 and 4 in Table 4 show that external CEOs who

work for family �rms spend 32% more time with insiders and 27% less time with outsiders

alone, but only the former is estimated precisely. This is in line with the intuition that,

compared to disperse shareholders, family owners can keep a tighter control on their CEOs.

Family CEOs, in contrast, do not appear to behave di¤erently than CEOs in non family

�rms.

The second indicator is the size of the board. Large boards have long been suspected

of being dysfunctional and to provide too little criticism of management performance, thus

granting CEOs more discretion which can be abused (e.g. Lipton and Lorsch, 1992). Jensen

(1993), subscribing to this criticism argues that "�when boards get beyond seven or eight

people they are less likely to function e¤ectively and are easier for the CEO to control". In

addition, pletoric boards can be just the re�ection of attempts to bring in friends and transfer

them shareholders money through board members compensation. Consistent with this views

there is some evidence that larger boards reduce the value of the �rm (e.g. Yermack, 1996).

In line with this, Columns 7 and 8 show that CEOs with larger boards spend less time with

insiders and more time with outsiders alone. A 1% percentage point increase in board size

reduces the time spent with insiders by 0.24 of a percentage point while it increases the time

spent with outsiders alone by 0.47 of a percentage point.

A better measure of constraints on the CEO is the independence of board members. We

have no direct information on the nature of the directors. But we observe some of their

characteristics, including gender. As a third proxy for quality of governance we use an

indicator for whether there is at least a woman in the board. There is evidence that the

presence of women improves the quality of governance. For instance, Adams and Ferreira

(2009) �nd that boards with more women tend to be tougher with the CEO, to raise board

attendance and to monitor more.7 Throughout the analysis, we distinguish between executive

7One rational is that women have higher standards of morality and a higher sense of their duty as docu-

mented in experimental and survey based studies which show that women are more likely to stick to the rules,

less likely to break agreed legal or moral norms (e.g. Guiso et. al (2003) ) and are less self-oriented then men

(for instance Eagly and Crowley (1986); Eckel and Grossman (1998); Reiss and Mitra (1998); Butler et. al.

17

and non executive roles, since we expect this e¤ect to be stronger when women assume

executive roles. The �ndings in columns 9 and 10 indicate that CEOs spend 17% more time

with insiders and 27% less with outsiders alone, respectively when at least one woman on the

board has an executive role. As expected, the correlations have a similar sign but are smaller

in magnitude and not precisely estimated when we look at the presence of women with non

executive roles.

As a �nal and more direct measure of governance quality and restraints on the man-

agers we use the country-level indicator of private bene�ts of control computed by Dyck and

Zingales (2004) on the basis of block control transactions in 39 countries. We attach this

measure of quality of governance to multinational �rms in our sample on the basis of their

country of origin and rely on within variation among multinational �rms operating in Italy.

Compared to the others, the main advantages of this measure are its direct interpretability in

terms of access to private bene�ts and its exogeneity. It comes at the cost that we can only

rely on a subset of observations. The �ndings in columns 11 and 12, table 4, indicate that

CEOs who work for multinationals based in countries with better governance spend more

time with insiders and less with outsiders alone. A one standard deviation increase in the

index, indicating a worsening of governance, decreases hours spent with insiders by 13% and

increases hours spent with outsiders alone by 45%.

Taken together, the �ndings in Table 4 provide a coherent picture. Whenever the CEO

and the �rm interest are better aligned - either because of stronger market competition,

better monitoring, more powerful boards- CEOs spend more time with insiders and less time

with outsiders alone. This, as well as the �ndings in the previous section, suggests that

the evidence is consistent with the idea that time with outsiders alone yields more private

bene�ts than time with insiders, and that di¤erences in time allocation re�ect di¤erences

in governance. The next section looks at whether di¤erences in time allocation have any

implication for �rm performance.

5.3 CEOs�Time Use and Firm Performance

For our next test we match the time use data with external measures of �rm performance

from the Amadeus database. Following the theory, we test whether the time the CEO spends

with di¤erent categories of people is correlated with �rm performance. As illustrated by the

model in Section 3, all hours devoted to productive activities should be positively correlated

with �rm performance, namely the CEOs�time is an input in the production function. If,

on the other hand, the CEO spends time on activities that bene�t him but not the �rm, the

(2010)). Because of this they are more willing to voice in the board meetings against a shirking CEO, thus

providing more monitoring and control.

18

time devoted to these activities should not be correlated to �rm performance.

Table 5 estimates the conditional correlation between �rm performance and CEOs�time

use controlling for �rm size and industry dummies. Our main measure of performance is

productivity, measured as the value of sales over employees, which is available for 80 of

our 94 sample �rms. Firms in the �nance sector are excluded from this exercise because

sales/employees is not a comparable measure of productivity in this sector.

Column 1 shows a strong positive correlation between hours worked and productivity.

The coe¢ cient is precisely estimated and implies that a one percentage point increase in

CEOs�working hours is associated with a 2.14 percentage point increase in productivity.

Column 2 shows that this correlation is entirely driven by the hours the CEOs spend with

at least one other employee of the �rm. A one percentage point increase in hours spent with

at least one insider is associated with a 1.23 percentage points increase in productivity. In

contrast, hours spent working alone and, most interestingly, hours spent with outsiders alone

are not correlated with productivity. The coe¢ cient is precisely estimated but small and not

signi�cantly di¤erent from zero. The test of equality of coe¢ cients between hours with at

least one insiders and hours with outsiders alone rejects the null at the 1% level.

Columns 3 to 5 address measurement error issues. In particular, given that the CEO

generally meets insiders at the �rm�s HQ, while he might meet outsiders in other locations,

the results could be driven by measurement error if the time spent with outsiders comprises

unproductive travel or setting-up time. Alternatively, the PA might have more precise in-

formation on the hours dedicated to a given activity if this takes place in the CEO�s o¢ ce,

which is presumably close to hers, rather than on activities that take place away from the

�rm. To allay this concern, Column 3 splits the hours spent with outsiders alone according

to location. If this were driving the results, we would expect the time spent with outsiders at

HQ to be positively correlated with �rm productivity. The results indicate that however this

is not the case. The coe¢ cient on hours spent with outsiders is small, precisely estimated

and not signi�cantly di¤erent from zero regardless of where the activity takes place.

A related concern is that the PA might have more precise information about the duration

of activities with insiders and hence measurement error would again introduce a downward

bias on the coe¢ cient on time with outsiders alone. We address this concern using infor-

mation on whether the activity was scheduled in advance and recorded in the CEO�s diary.

Intuitively, activities that are planned in advance and already recorded should be measured

more precisely. Columns 4 and 5 show that this form of measurement error was not driving

the results either. The coe¢ cient of hours spent with outsiders is small, precisely estimated

and not signi�cantly di¤erent from zero regardless of whether the activity was planned in

advance. In contrast, the coe¢ cient on hours spent with at least one insider is positive and

19

signi�cantly di¤erent from zero regardless of whether the activity was planned in advance.

Column 5 probes the robustness of the results to the inclusion of other inputs, namely capital

and materials.

While the pattern of time use is correlated with productivity, it might also be correlated

with other unobservables - most notably the CEOs pay - so to leave the �rm owners indif-

ferent between having a hard working CEO who spends long hours with insiders and a CEO

who works fewer hours and spends more time with outsiders alone. To assess the empirical

relevance of this interpretation, Column 6 presents the conditional correlation between �rms

pro�ts per employee and time use. The results show that time spent with at least one insider

is positively correlated with pro�ts, whereas time spent with outsiders is not.

A possible further concern is that the impact of time allocation choices might reveal

itself over a longer time period. For example, networking activities might need a longer

time frame to bear their positive impact on �rm performance. To test this idea we consider

as a dependent variable the growth rate of sales over a three year period. The results in

column 7 do not support the idea that time spent with outsiders might have a positive long

run impact. In fact, although the coe¢ cients are not precisely estimated, time spent with

outsiders is negatively correlated with three-year sales growth, while time spent with insiders

shows a positive correlation.

6 Discussion and Conclusions

The analysis above has illustrated a rich set of correlations between the CEOs�pattern of time

use, his e¤ort, governance proxies and the performance of the �rm he leads. In particular we

�nd that CEOs who work longer hours spend more time in activities that involve at least one

insider and less time in activities where the CEO meets outsiders alone. CEOs who work for

�rms with stronger governance also spend more time in activities that involve at least one

insider and less time in activities where the CEO meets outsiders alone Finally, time spent

with insiders is positively correlated with productivity and pro�ts, whereas time spent with

outsiders alone is not.

As illustrated in the theoretical framework, variations in time use can either be seen as

optimal responses to di¤erent circumstances faced by di¤erent �rms if the CEO�s ad the

�rm�s interests are perfectly aligned (b = 1) or as consequences of di¤erences in the quality of

governance. We now make precise the conditions under which the �ndings can be reconciled

with either the perfect and the imperfect governance view. It is important to stress that our

data does not allow us to implement a formal test of the two models, but the rich set of

correlations we uncover allows us to pin down the set of assumptions needed to reconcile the

20

�ndings with either view.

In the perfect governance view, where the CEO�s and the �rm�s interests are perfectly

aligned, the observed correlations must be due to shocks that a¤ect the pattern of time use,

total hours worked, �rm performance and �rm characteristics as follows. First, shocks that

increase the marginal return to CEO�s e¤ort - so that CEOs work longer hours, also increase

the marginal product of time with at least one insider and at the same time decrease the

marginal product of time alone with outsiders, especially so for time alone in one-to-one

meetings. Second, shocks that increase the �rm�s productivity, also increase the marginal

product of time with at least one insider and leave the marginal product of time alone with

outsiders constant. Third, this type of shocks are more likely to hit multinationals, family

�rms with external CEOs, �rms with small boards, �rms with women covering executive roles

on the board, and �rms from countries with a good governance record.

In the imperfect governance view, the observed variation in time use can derive from

di¤erences in governance that determine the extent to which the CEOs�interests are aligned

with the �rm. If we take this view, the �ndings indicate that time spent with insiders

contributes to the �rm�s output while time spent alone with outsiders only bene�ts the

CEO. Moreover, CEOs hired by �rms with better governance work longer hours and devote

more time to activities that bene�t the �rm and less time to activities that mostly carry

private bene�ts.

We leave it to the reader to assess which of the two interpretations is more plausible. In

either case, our analysis suggests that CEO time use data are a useful tool to shed light on

constraints and objectives of �rms and of their executives.

References

[1] Adams, Renée B. and Ferreira, Daniel. 2009. "Women in the Boardroom and Their

Impact on Governance and Performance." Journal of Financial Economics, 94 (2):291-

309.

[2] Aguiar, Mark and Erik Hurst. 2007. Measuring Trends in Leisure: The Allocation of

Time Over Five Decades 2007. Quarterly Journal of Economics, August 2007, 122(3),

969-1006

[3] Bandiera, Oriana; Guiso, Luigi; Prat, Andrea and Sadun, Ra¤aella. 2010. "Matching

Firms, Managers, and Incentives." Harvard Business School Working Paper 10-073.

http://www.hbs.edu/research/pdf/10-073.pdf.

21

[4] Bennedsen, Morten; Nielsen, Kasper Meisner; Perez-Gonzalez, Francisco and Wolfenzon,

Daniel. 2007. �Inside the Family Firm: The Role of Families in Succession Decisions and

Performance.�Quarterly Journal of Economics, 122:647-91.

[5] Bertrand, Marianne and Schoar, Antoinette. 2003. "Managing with Style: The E¤ect of

Managers on Firm Policies." Quarterly Journal of Economics, 118 (4):1169-208.

[6] Bloom, Nicholas and Van Reenen, John. 2010. "Why Do Management Practices Di¤er

across Firms and Countries?" Journal of Economic Perspectives, 24 (1):203-24.

[7] Bolton, Patrick and Dewatripont, Mathias. 1994. "The Firm as a Communication Net-

work." The Quarterly Journal of Economics, 109 (4):809-39.

[8] Butler, Je¤ ; Giuliano, Paola and Guiso, Luigi. 2010. "Trust and Cheating." UCLA

mimeo.

[9] Drucker, Peter F. 1966. The E¤ective Executive. New York,: Harper & Row.

[10] Drucker, Peter F.; D�Antonio, William V. and Ehrlich, Howard J. 1961. Power and

Democracy in America. Notre Dame, IN: University of Notre Dame Press.

[11] Dyck, Alexander and Zingales, Luigi. 2004. "Private Bene�ts of Control: An Interna-

tional Comparison." The Journal of Finance, 59 (2):537-600.

[12] Eagly, A., and M. Crowley. 1986. "Gender and Helping Behavior: A Meta-analytic View

of the Social Psychological Literature." Psychological Bulletin 100:283-308.

[13] Eckel, Catherine C. and Grossman, Philip J. 1998. "Are Women Less Sel�sh Than Men?:

Evidence from Dictator Experiments." Economic Journal, 108 (448):726-35.

[14] Garicano, Luis. 2000. "Hierarchies and the Organization of Knowledge in Production."

Journal of Political Economy, 108 (5):874-904.

[15] Guiso, Luigi; Sapienza, Paola and Zingales, Luigi. 2003. "People�s Opium? Religion and

Economic Attitudes." Journal of Monetary Economics, 50 (1):225-82.

[16] Hamermesh Daniel.1990. "Shirking or Productive Schmoozing: Wages and the Alloca-

tion of Time at Work," Industrial and Labor Relations Review, January

[17] Jensen, Michael C. 1993. "The Modern Industrial Revolution, Exit, and the Failure of

Internal Control Systems." The Journal of Finance, 48 (3):831-80.

[18] Khurana, Rakesh. 2002. Searching for a Corporate Savior: The Irrational Quest for

Charismatic CEOs. Princeton, NJ: Princeton University Press.

22

[19] Kotter, John P."�What E¤ective General Managers Really Do." Harvard Business Re-

view, March-April 1999, p. 145-159.

[20] Krueger, Alan B. and Andreas Mueller. 2010. "Job Search and Unemployment Insurance:

New Evidence from Time Use Data," Journal of Public Economics, forthcoming.

[21] Krueger, Alan B., Daniel Kahneman; David Schkade; Norbert Schwarz and Arthur A.

Stone. 2009. "National Time Accounting: The Currency of Life," Measuring the Sub-

jective Well-Being of Nations, University of Chicago Press for NBER, 2009, pp. 9-86.

[22] La�ey, A. G. 2009. "What Only the CEO Can Do." Harvard Business Review, 87 (5):54-

62.

[23] Lipton, Martin and Lorsch, Jay W. 1992. "A Modest Proposal for Improved Corporate

Governance." Business Lawyer, 48 (1):59-77.

[24] Luthans, Fred. 1988. "Successful vs. E¤ective Real Managers." The Academy of Man-

agement Executive (1987-1989), 2 (2):127-32.

[25] Malmendier, Ulrike and Tate, Geo¤rey. 2009. "Superstar CEO�s." Quarterly Journal of

Economics, 124 (4):1593-638.

[26] Mintzberg, Henry. 1973. The nature of managerial work. New York,: Harper & Row.

[27] Pérez-González, Francisco. 2006. �Inherited Control and Firm Performance.�American

Economic Review, 96:1559-88.

[28] Reiss, Michelle C. and Mitra, Kaushik. 1998. "The E¤ects of Individual Di¤erence Fac-

tors on the Acceptability of Ethical and Unethical Workplace Behaviors." Journal of

Business Ethics, 17 (14):1581-93.

[29] Stein, Jeremy C. 2003. "Agency, Information and Corporate Investment" in Handbook

of the Economics of Finance, eds. G. M. Constantinides, M. Harris and R. M. Stulz,

109-63. Amsterdam; Boston: Elsevier/North-Holland.

[30] Yermack, David. 1996. "Higher Market Valuation of Companies with a Small Board of

Directors." Journal of Financial Economics, 40 (2):185-211.

[31] Van Zandt, Timothy. 1998. "The Scheduling and Organization of Periodic Associative

Computation: E¢ cient Networks." Review of Economic Design, 3 (2):93-127.

23

FIGURE 1: CEO's Time Use1.A Share of time devoted to different activities

1B Sh f ti t ith diff t t i f l

mean of sharetime_lunch

mean of sharetime_publicevent

mean of sharetime_phonecalls

mean of sharetime_workalone

mean of sharetime_meetings

0 .2 .4 .6

meetings

works alone

phone calls

public events

business lunches

1B. Share of time spent with different categories of people

mean of sharetime_lunch

mean of sharetime_publicevent

mean of sharetime_phonecalls

mean of sharetime_workalone

mean of sharetime_meetings

0 .2 .4 .6

meetings

works alone

phone calls

public events

business lunches

insiders and outsiders

insiders and outsidersmean of so

mean of sharetime_mix

mean of si

0 .1 .2 .3 .4

outsiders alone

insiders only

insiders and outsiders

FIGURE 2: CEO's Time Use

2A. Hours Worked

2B. Hours spent with outsiders alone

1.2

.3.4

Fra

ctio

n0

.05

.1.1

5.2

.25

Fra

ctio

n

20 30 40 50 60weekly hrs all activitiestotal weekly hours

2C. Hours spent with at least one insider

0.1

.2.3

.4F

ract

ion

0 5 10 15 20weekly hrs spent in this activity

0.0

5.1

.15

.2.2

5F

ract

ion

0 10 20 30 40 50weekly hrs spent in this activity

0.0

5.1

.15

.2.2

5F

ract

ion

20 30 40 50 60weekly hrs all activitiestotal weekly hours

weekly hours with outsiders alone

weekly hours with at least one insider

Note: Weekly hours are computed as the sum of time spent in all activities longer thanfifteen minutes. Weekly hours with outsiders alone is the sum of time spent in all activitieswhere the CEO was alone with outsiders for longer than fifteen minutes. Weekly hourswith outsiders alone is the sum of time spent in all activities where the CEO was with at

Table 1 Sample Descriptives

Mean Median Std. Dev. ObsSize (Number of Employees) 8330 1244 30014 94Productivity (Sales/Employees X1000 USD) 4234 600 13664 80Sector: (%) 98

Manufacturing 39.4Services 13.8

Finance, Insurance, Real Estate 14.9Transportation and Communication 12.8

Wholesale and Retail 8.5Mining 3.2

Construction 4.2Ownership (%) 94

Widely Held 37.2Family 22.3State 18.1Other 13.3

Nationality (%) 98Domestic 29.9

Italian MNE 36.1 .Foreign MNE 34

Firm is listed 0.19 94Board size 13.1 10 8.1 91At least one woman on the board in an executive role 0.56 91At least one woman on the board in a non executive role 0.68 91Private benefit index 0.023 0.038 0.049 30

.

Data Sources and Variable Definitions: Data on firm size, productivity and sector are drawn from AMADEUS. Productivity isnot available for firms in the financial sector. Ownership and multinational status has been obtained through manualsearch. Board data is drawn from the Infocamere dataset. Accounting, ownership and board information are relative to2007. The private benefit index is measured (only for non Italian multinationals) at the country of origin level and it is drawnfrom Dyck and Zingales (2004), Table V, col 2.

TABLE 2: CEOs' Time Use

AverageShare of CEOs with

hrs>0

Total Hours 47.7

(9.98)

Total Hours Recorded 36.94

(8.33)Working alone 5.66 .91

(4.65)With at least one insider 24.2 1

(9.35)With outsiders only 4.88 .84

(4.83)

Insiders, Top 5 Categories:

Finance 8.69 .79(6.96)

Marketing 8.32 .73(7.62)

BU Directors 6.85 .71(7.18)( )

Strategy 5.87 .69(5.97)

Human Resources 5.47 .74(5.86)

Outsiders, Top 5 Categories

Clients 2.59 .40(4.03)

Suppliers 1.27 .34(2.19)

Investors 1.74 .33(3.05)

Consultants 4.70 .67(4.82)

Banks 1.38 .30(2.68)

Data Sources and Variable Definitions: All variables are drawn from the time use survey. Totalhours include all activities, including those shorter than 15 minutes and travel. Total hoursrecorded only contains activities that are longer than 15 minutes and does not include travel. Thefirst column reports average hours spent by the sample CEOs over the relevation week with eachof the listed categories in activities longer than 15 minutes. The second column reports the shareof CEOs that spend at least 15 minutes with each of the listed categories.

TABLE 3. CEOs' Time Use and Hours WorkedPanel A Insiders and Outsiders

1 2

log(total hours) spent: with at least one insider with outsiders only

log (total hours) 1.179*** 0.693**(0.147) (0.337)

H0: log(total hours)=1 .23 .00

Adjusted R squared 0.473 0.207Controls size, listed size, listedIndustry Dummies yes yesObservations 94 94

Panel B Insiders and Outsiders, by number of categories involved

3 4 5 6

log(total hours) spent:with one category of

insiderswith more than onecategory of insiders

with one category ofousiders

with more than onecategory of outsiders

log (total hours) 0.909*** 1.082** 0.926*** 0.381*(0.260) (0.448) (0.317) (0.223)

log (number of employees) 0.024 0.028 0.053 0.029(0.038) (0.054) (0.046) (0.027)

listed firm (=1 if yes) 0.164 0.631** 0.906*** 0.146(0.255) (0.287) (0.214) (0.202)

H0: log(total hours)=1 .73 .85 .00 .00

d d dAdjusted R squared 0.138 0.253 0.205 0.166Controls size, listed size, listed size, listed size, listedIndustry Dummies yes yes yes yesObservations 94 94 94 94

Panel C Insiders and Outsiders, by number of participants involved

7 8 9 10

log(total hours) spent:with fewer than two

participants and at leastone insider

with more than twoparticipants and atleast one insider

with outsiders only andfewer than twoparticipants

with outsiders onlyand more than two

participants

log (total hours) 0.594* 1.387*** 0.789*** 0.075(0.303) (0.224) (0.276) (0.365)

log (number of employees) 0.010 0.027 0.040 0.013(0.040) (0.032) (0.037) (0.039)

listed firm (=1 if yes) 0.082 0.182 0.566** 0.640**(0.201) (0.130) (0.251) (0.248)

H0: log(total hours)=1 .18 .08 .00 .00

Adjusted R squared 0.136 0.422 0.168 0.142Controls size, listed size, listed size, listed size, listedIndustry Dummies yes yes yes yesObservations 94 94 94 94

Notes: *,**,*** indicate that the coefficient is significantly different from zero at the 10%, 5%, 1% level, respectively. OLS Estimates, robuststandard errors in parentheses. All regressions include industry dummies, the logarithm of the number of employees and a dummy equal to 1if the firm is listed.

TABLE 4. CEOs' Time Use and Governance

1 2 3 4 5 6 7 8 9 10 11 12

dependent variable is log(total hours) spent with:with atleast oneinsider

outsidersonly

at leastone insider

outsidersonly

at leastone insider

outsidersonly

at leastone insider

outsidersonly

at leastone insider

outsidersonly

at leastone insider

outsidersonly

Foreign MNE 0.428*** 0.594** 0.438*** 0.612**(0.123) (0.257) (0.125) (0.260)

Domestic MNE 0.288** 0.233 0.217* 0.122(0.122) (0.227) (0.124) (0.233)

Family owned External CEO 0.277** 0.337 0.313** 0.486(0.125) (0.293) (0.135) (0.301)

Family owned Family CEO 0.115 0.135 0.072 0.046(0.106) (0.272) (0.118) (0.288)

Log (board size) 0.243* 0.474**(0.135) (0.230)

At least one woman on the board in an executive role 0.165* 0.328*(0.096) (0.190)

At least one woman on the board in a non executive role 0.032 0.250(0.108) (0.241)

Private benefit index 2.605* 9.095*(1.152) (4.266)

Industry Dummies yes yes yes yes yes yes yes yes yes yes yes yesFirm Size yes yes yes yes yes yes yes yes yes yes yes yesFirm Listed yes yes yes yes yes yes yes yes yes yes yes yesAdjusted R squared 0.199 0.230 0.115 0.190 0.243 0.253 0.128 0.218 0.103 0.206 0.291 0.299Observations 94 94 94 94 94 94 88 88 88 88 30 30

Notes: *,**,*** indicate that the coefficient is significantly different from zero at the 10%, 5%, 1% level, respectively. OLS Estimates, robust standard errors in parentheses. All regressions include industry dummies, the logarithmemployees and a dummy equal to 1 if the firm is listed. Sample size in columns 7 10 falls due to missing values of the board variables. The private benefit index is the country level indicator of private benefits of control comput(2004) on the basis of block control transactions in 39 countries., which we match to the country of origin of the foreign multinationals in our sample. Consequently, the sample in columns 11 14 is restricted to foreign multinatio

Table 5: CEOs' Time Use and Firm Performance

(1) (2) (2) (3) (4) (5) (7) (8)

dependent variable: Log(productivity) Log(productivity) Log(productivity) Log(productivity) Log(productivity) Log(productivity) Profits/Employees Sales growth rate

Log total hours 2.145***(0.611)

Log total hours with at least one insider 1.229*** 1.274*** 1.223*** 0.740** 72.340* 0.071(0.431) (0.413) (0.418) (0.319) (41.758) (0.077)

Log total hours with outsiders only 0.227 0.072 31.758 0.016(0.249) (0.189) (21.187) (0.061)

Log total hours working alone 0.258 0.278 0.284* 0.237 0.166 3.619 0.005(0.162) (0.168) (0.161) (0.157) (0.152) (13.439) (0.046)

Log total hours with outsiders at headquarters 0.326(0.294)

Log total hours with outsiders outside 0.114(0.152)

Log total hours with outsiders planned in advance 0.210 0.205(0.215) (0.204)

Log total hours with outsiders unplanned 0.196 0.219(0.342) (0.347)

Log total hours with insiders planned in advance 0.857***(0 323)(0.323)

Log total hours with insiders unplanned 0.340*(0.186)

H0: hours with insiders=hours with outsiders .000 .006 .184 .244H0: hours with outsiders at HQ=hours with outsiders outside .421H0: hours with insiders planned=hours with insiders unplanned .131H0: hours with outsiders planned=hours with outsiders unplanned 0.288 .247Industry Dummies yes yes yes yes yes yes yes yes

Controls size , listed size , listed size , listed size , listed size , listedsize , listed, log

(capital),log(materials)

size , listed size , listed

Adjusted R squared 0.480 0.476 0.491 0.483 0.478 0.526 0.476 0.111Observations 80 80 80 80 80 78 76 66

Notes: *,**,*** indicate that the coefficient is significantly different from zero at the 10%, 5%, 1% level, respectively. OLS Estimates, robust standard errors in parentheses.All regressions include industry dummies, the logarithm of the number of employees and adummy equal to 1 if the firm is listed. Productivity is defined as sales/employees. Sales growth rate is computed as the logharitmic change over a three year period.

Figure A1: CEO Fixed Effects with 95% Confidence Interval

-20

24

(mean) DFE_I-6

-4-2

02

4(mean) DFE_I

cid

Note: Each dot represents the coefficient on each CEO dummy in a regression of log(timewith at least one insider) on firm size, industry dummies and CEO dummies at the activitylevel. The bars are the 95% confidence intervals.

TABLE A1. Selection into the sample.

1 2 3 4 5dependent variable:

log (number of employees) 0.020** 0.026** 0.026** 0.022** 0.030**(0.009) (0.010) (0.011) (0.011) (0.012)

Part of "Amici Fondazione De Benedetti" list (=1 if yes 0.373*** 0.251** 0.250** 0.271** 0.443***(0.101) (0.108) (0.109) (0.110) (0.132)

MNE firm (1=yes) 0.038 0.038 0.047 0.055(0.038) (0.038) (0.038) (0.040)

listed firm (=1 if yes) 0.082 0.082 0.107 0.039(0.075) (0.076) (0.079) (0.083)

Geographical Area: Islands 0.003 0.012 0.004 0.161(0.112) (0.124) (0.115) (0.167)

Geographical Area: North East 0.010 0.010 0.009 0.041(0.044) (0.045) (0.045) (0.047)

Geographical Area: North West 0.057 0.057 0.060 0.053(0.041) (0.041) (0.041) (0.043)

Geographical Area: South 0.024 0.032 0.023 0.007(0.063) (0.065) (0.062) (0.076)

Finance, Insurance 0.187*** 0.190*** 0.199*** 0.171**(0.062) (0.063) (0.065) (0.073)

Manufacturing 0.029 0.030 0.030 0.012(0.036) (0.037) (0.036) (0.039)

Mining 0.729*** 0.725*** 0.740*** 0.349***(0.193) (0.192) (0.209) (0.134)

Public Administration 0.637*** 0.639*** 0.602*** 0.482***(0.116) (0.118) (0.119) (0.146)

Retail Trade 0.002 0.002 0.001 0.008(0.077) (0.076) (0.078) (0.073)

Services 0.289*** 0.290*** 0.284*** 0.337***(0.092) (0.095) (0.093) (0.110)

Transportation 0.005 0.005 0.018 0.019(0.045) (0.045) (0.046) (0.049)

log(sales) 0.050(11.431)

Profits/Employees (millions) 0.041(0.070)

Sales Growth Rate (3 years) 0.076(0.054)

Constant 0.030 0.111 0.114 0.091 0.127(0.054) (0.073) (0.160) (0.081) (0.085)

Adjusted R squared 0.0614 0.1432 0.1429 0.1476 0.1847Observations 535 523 521 515 422

Dummy=1 if firm is in sample

Notes: *, **, *** indicated that the coefficient is significantly different from zero at the 10%, 5% and 1% level, respectively. OLS Estimates,robust standard errors in parentheses. The omitted category for geographical area is "Centre". The omitted category for sector is "construction".